Skip to content

Navigation Menu

Appearance settings

- AI CODE CREATION
  GitHub CopilotWrite better code with AI
  GitHub Copilot appDirect agents from issue to merge
  MCP Registry^NewIntegrate external tools
- DEVELOPER WORKFLOWS
  ActionsAutomate any workflow
  CodespacesInstant dev environments
  IssuesPlan and track work
  Code ReviewManage code changes
- APPLICATION SECURITY
  GitHub Advanced SecurityFind and fix vulnerabilities
  Code securitySecure your code as you build
  Secret protectionStop leaks before they start
- EXPLORE
  Why GitHub
  Documentation
  Blog
  Changelog
  Marketplace
View all features
- BY COMPANY SIZE
  Enterprises
  Small and medium teams
  Startups
  Nonprofits
- BY USE CASE
  App Modernization
  DevSecOps
  DevOps
  CI/CD
  View all use cases
- BY INDUSTRY
  Healthcare
  Financial services
  Manufacturing
  Government
  View all industries
View all solutions
- EXPLORE BY TOPIC
  AI
  Software Development
  DevOps
  Security
  View all topics
- EXPLORE BY TYPE
  Customer stories
  Events & webinars
  Ebooks & reports
  Business insights
  GitHub Skills
- SUPPORT & SERVICES
  Documentation
  Customer support
  Community forum
  Trust center
  Partners
View all resources
- COMMUNITY
  GitHub SponsorsFund open source developers
- PROGRAMS
  Security Lab
  Maintainer Community
  Accelerator
  GitHub Stars
  Archive Program
- REPOSITORIES
  Topics
  Trending
  Collections
- ENTERPRISE SOLUTIONS
  Enterprise platformAI-powered developer platform
- AVAILABLE ADD-ONS
  GitHub Advanced SecurityEnterprise-grade security features
  Copilot for BusinessEnterprise-grade AI features
  Premium SupportEnterprise-grade 24/7 support
Pricing

Search code, repositories, users, issues, pull requests...

Search

Clear

Search syntax tips

Provide feedback

We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Saved searches

Use saved searches to filter your results more quickly

Name

Query

To see all available qualifiers, see our documentation.

Appearance settings

You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

Uh oh!

There was an error while loading. Please reload this page.

cornball-ai / diffuseR Public

Notifications You must be signed in to change notification settings
Fork 3
Star 47

Code
Issues 1
Pull requests 3
Actions
Projects
Security and quality
Insights

Additional navigation options

Code
Issues
Pull requests
Actions
Projects
Security and quality
Insights

FLUX.2 Klein 4B text-to-image support#21

Merged

TroyHernandez merged 84 commits into

maincornball-ai/diffuseR:mainfrom

feature/flux2-kleincornball-ai/diffuseR:feature/flux2-kleinCopy head branch name to clipboard

Jul 7, 2026

Conversation Commits84 (84)Checks Files changed

Merged

FLUX.2 Klein 4B text-to-image support#21
TroyHernandez merged 84 commits into
mainfrom
feature/flux2-klein

Commits

Commits on Jul 3, 2026

Remove LTX-2.0 implementation ahead of clean-room LTX-2.3 rewrite
TroyHernandez
committed
License hygiene: attribution, ignores, CLAUDE.md provenance rewrite
TroyHernandez
committed
Add LTX-2.3 single-file checkpoint reader
TroyHernandez
committed
Port LTX-2.3 rotary position embeddings from diffusers
TroyHernandez
committed
Port LTX-2.3 audio-video DiT from diffusers
TroyHernandez
committed
Port LTX-2.3 text connectors from diffusers
TroyHernandez
committed
Port LTX-2.3 causal video VAE from diffusers
TroyHernandez
committed
Port LTX-2.3 audio VAE decoder and BWE vocoder from diffusers
TroyHernandez
committed
Add LTX-2.3 distilled text-to-video pipeline
TroyHernandez
committed
Add fp8 quantization, memory profiles, and download UX
TroyHernandez
committed
Add spatial latent upsampler and two-stage generation
TroyHernandez
committed
Add treesitR coverage report for the LTX-2.3 port
TroyHernandez
committed
Handle bfloat16 in the connector mask (torch_finfo fallback)
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Fix R CMD check warnings
TroyHernandez
committed
Phase-sequential component offloading
TroyHernandez
committed
Add NF4 quantization with a GPU-resident transformer profile
TroyHernandez
committed
Encode precision strategy in the LTX-2.3 memory profiles
TroyHernandez
committed
Reuse NF4 dequantization buffers to eliminate allocator churn
TroyHernandez
committed
Make NF4 dequantization fully in-place over persistent scratch
TroyHernandez
committed
Add per-block VRAM tracing behind diffuseR.debug
TroyHernandez
committed

Commits on Jul 4, 2026

Extend phase offloading to the transformer
TroyHernandez
committed
Adaptive attention chunking and per-step timing
TroyHernandez
committed
Buffered attention: score/context matrices in persistent scratch
TroyHernandez
committed
Validated squeeze wrap-up: step timing, expandable segments, honest caps
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Tiled VAE decoding and corrected allocator tuning
TroyHernandez
committed
Document technique provenance in inst/REFERENCES.md
TroyHernandez
committed
In-place chunked tanh GELU for large feed-forward activations
TroyHernandez
committed
Default attn budget 1.5e8; raise NF4 profile cap to 1280x704
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Match HF hidden-state semantics in the Gemma3 encoder
TroyHernandez
committed
Remove committed Python venv from inst/validation
TroyHernandez
committed
Fix R CMD check findings: Rd percent escape, codetools-visible fn
TroyHernandez
committed
Bump version to 0.1.0
TroyHernandez
committed
TorchScript-compiled NF4 block stack
TroyHernandez
committed
Fix JIT stack for real-model layouts; 11.4x measured
TroyHernandez
committed
Reference the JIT block-stack provenance
TroyHernandez
committed
rformat + document
TroyHernandez
committed

Commits on Jul 7, 2026

JIT-traced VAE decode and audio chain
TroyHernandez
committed
Traced decode (opt-in), allocator findings, tune_gc VRAM fallback
TroyHernandez
committed
Apply the full mlverse allocator knob set; per-format allocator default
TroyHernandez
committed
FLUX RoPE, position ids, and latent pack/unpack (parity-tested)
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX MMDiT blocks: adaLN variants, joint attention, double/single blocks (parity-tested)
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX transformer assembly with parity fixture; tensor-method scalar adds
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX checkpoint readers and NF4/fp8 quantize + format-dispatching loader
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Unigram (T5/SentencePiece) tokenizer with Viterbi segmentation
TroyHernandez
committed
rformat + document
TroyHernandez
committed
T5-v1.1 encoder: relative position bias, gated-GELU, sharded loader
TroyHernandez
committed
rformat + document
TroyHernandez
committed
CLIP pooled output + safetensors loaders; parameterize VAE decoder channels
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX.1-schnell pipeline: loader, denoise loop, download flow, memory profiles
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Check hygiene: unicode escapes, fixture metadata, flux_quantize output_dir default
TroyHernandez
committed
treesitR structural translation audit tool
TroyHernandez
committed
Tokenizer parity vs shipped artifact: prepend-always cases, empty-input fix
TroyHernandez
committed
Quantized residents follow the compute dtype (fp32 on CPU)
TroyHernandez
committed
CLAUDE.md: FLUX.1-schnell section with measured numbers
TroyHernandez
committed
FLUX.2 position ids, 4-axis RoPE reuse, latent patchify chain, empirical mu
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX.2 blocks: shared modulation, SwiGLU FF, parallel single block
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX.2 transformer assembly; bias params on shared timestep/norm modules
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX.2 quantization: family dispatch via config class, resident fp8
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Qwen2 byte-level BPE tokenizer with Qwen3 chat template
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Qwen3 text encoder: GQA, q/k norms, padding mask, mid-stack states
TroyHernandez
committed
rformat + document
TroyHernandez
committed
FLUX.2 VAE decoder: post_quant_conv, BN latent statistics, decoder reuse
TroyHernandez
committed
FLUX.2 klein pipeline: loader, denoise loop, download flow, dispatcher
TroyHernandez
committed
rformat + document
TroyHernandez
committed
Extend treesitR audit with FLUX.2 and Qwen3 pairings
TroyHernandez
committed
REFERENCES.md: FLUX.2 klein provenance
TroyHernandez
committed
Perf: integer-id BPE (gc 203ms->19ms), buffered fp8 dequant, allocator tuning
TroyHernandez
committed
CLAUDE.md: FLUX.2 klein section with measured numbers
TroyHernandez
committed
Merge main (FLUX.1 squash) into feature/flux2-klein
TroyHernandez
committed
Bump version to 0.1.0.2
TroyHernandez
committed

Footer

© 2026 GitHub, Inc.

Footer navigation

Terms
Privacy
Security
Status
Community
Docs
Contact

You can’t perform that action at this time.